Objective 2.4 Transform Numerical and Categorical Data AIP-210 Practice Questions (Page 2)

Questions 6–10

1. 6foundation · easy

   A categorical feature has missing values, and the categories are nominal with no inherent order. Which imputation strategy is most appropriate?

   • AImpute with the mean of the feature
   • BImpute with the median of the feature
   • CImpute with a random category
   • DImpute with the mode (most frequent category)
   Reveal Solution Discussion Select an answer first
2. 7foundation · easy

   Which type of machine learning algorithm is most sensitive to feature scaling, and therefore requires features to be scaled before training?

   • ARandom forest
   • BDecision trees
   • CK-nearest neighbors (KNN)
   • DLinear regression
   Reveal Solution Discussion Select an answer first
4. 8application · medium

   A data analyst is working on a dataset where the 'Age' feature has a non-linear relationship with the target variable 'InsurancePremium'. A linear regression model is being used, and the analyst wants to capture the non-linearity without adding polynomial terms. The team also wants to keep the model interpretable. Which approach should the analyst use?

   • AUse a log transformation on Age.
   • BApply min-max scaling to the Age feature.
   • CDiscretize Age into bins such as 18-25, 26-35, 36-50, 51-65, and 65+.
   • DReplace Age with its rank order.
   Reveal Solution Discussion Select an answer first
5. 9foundation · easy

   Which method is commonly used to detect outliers in a numerical feature by identifying values that fall below Q1 - 1.5 * IQR or above Q3 + 1.5 * IQR?

   • ABox plot method
   • BStandard deviation method
   • CZ-score method
   • DInterquartile range (IQR) method
   Reveal Solution Discussion Select an answer first
6. 10expert · hard

   A data scientist is building a logistic regression model to predict customer conversion. The dataset includes a 'DeviceID' feature with over 50,000 unique values, and the target conversion rate varies significantly across devices. The team wants to use this feature to improve model performance but is concerned about overfitting and data leakage. Which approach should the data scientist take?

   • AApply one-hot encoding to the DeviceID feature.
   • BApply label encoding to the DeviceID feature.
   • CApply target encoding with cross-validation to avoid leakage.
   • DApply frequency encoding to the DeviceID feature.
   Reveal Solution Discussion Select an answer first